Research map / Video / Action recognition
MiCT: Mixed 3D/2D Convolutional Tube for Human Action Recognition
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace stacked 3D convolutions with MiCT's mixed 2D/3D design | C3D · UCF101 video (35f, BN) | accuracy: 51.7% -> 58.7% | Helped |
| Replace stacked 3D convolutions with MiCT's mixed 2D/3D design | C3D · HMDB51 video (35f, BN) | accuracy: 48.0% -> 54.3% | Helped |
| Replace stacked 3D convolutions with MiCT's mixed 2D/3D design | C3D · Sport1M video (16f, BN) | accuracy: 60.8% -> 64.1% | Helped |
| Add concatenated 2D convolution connections (ablation) | MiCT-Net · UCF101 video (35f) | accuracy: 52.8% -> 56.1% | Helped |
| Add concatenated 2D convolution connections (ablation) | MiCT-Net · HMDB51 video (35f) | accuracy: 48.4% -> 51.9% | Helped |
| Add cross-domain residual connections (ablation) | MiCT-Net · UCF101 video (35f) | accuracy: 52.8% -> 58.7% | Helped |
| Add cross-domain residual connections (ablation) | MiCT-Net · HMDB51 video (35f) | accuracy: 48.4% -> 54.3% | Helped |
| Combine concatenated 2D connections and cross-domain residual connections (ablation) | MiCT-Net · UCF101 video (35f) | accuracy: 56.1% -> 58.7% | Helped |
| Combine concatenated 2D connections and cross-domain residual connections (ablation) | MiCT-Net · HMDB51 video (35f) | accuracy: 51.9% -> 54.3% | Helped |
| Initialize with ImageNet-pretrained weights (ablation) | MiCT-Net · UCF101 video (35f) | accuracy: 58.7% -> 87.3% | Helped |
| Initialize with ImageNet-pretrained weights (ablation) | MiCT-Net · HMDB51 video (35f) | accuracy: 54.3% -> 58.0% | Helped |
| Add an optical-flow stream and average predictions with the RGB stream | MiCT-Net · UCF101 | accuracy: 88.9% -> 94.7% | Helped |
Closest papers by meaning
- Deep Analysis of CNN-Based Spatio-Temporal Representations for Action Recognition (CVPR 2021 · similarity 0.91)
- T-C3D: Temporal Convolutional 3D Network for Real-Time Action Recognition (AAAI 2018 · similarity 0.91)
- PA3D: Pose-Action 3D Machine for Video Recognition (CVPR 2019 · similarity 0.88)
- Convolutional Two-Stream Network Fusion for Video Action Recognition (CVPR 2016 · similarity 0.88)
- ACTION-Net: Multipath Excitation for Action Recognition (CVPR 2021 · similarity 0.88)
- DVANet: Disentangling View and Action Features for Multi-View Action Recognition (AAAI 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.